A laser may be used as an optical transmitter that transmits light at a given wavelength. The power (i.e., amplitude) of the laser light may be modulated by corresponding modulation of the power used to drive the laser. In a directly-modulated electrically pumped semiconductor laser such as a laser diode, the electrical current that drives or pumps the laser is modulated. The relationship between the light output and the input current for such a laser may be represented using a transfer curve or L-I (light-current) curve. The set point of the L-I curve may be selected so as to maximize the linearity of the laser output in response to the modulation, within the expected range of operation of the output produced by the laser. Although the laser output may be generally linear along a significant portion of the L-I curve, the light output may attain a zero-power level when the input current falls below a threshold current level, which results in an effect known as clipping.
In a communications system where multiple channels are transmitted, such as a CATV system, multiple analog signals corresponding to the multiple channels may be combined into a wide-band multichannel RF signal, which drives a laser to produce a multichannel modulated optical signal. The multiple analog signals may include multiple modulated analog carriers that may be combined, for example, using frequency division multiplexing techniques. One or more digital signals modulated using digital modulation, such as quadrature amplitude modulated (QAM), may also be combined with the modulated analog carrier signals, for example, using subcarrier multiplexing (SCM) techniques. In some systems, for example, as many as 110 channels may be transmitted over a frequency range of about 50 MHz to 750 MHz.
Because the modulation may carry several channels of information at different frequencies, there may be a very large swing of the input drive current in either direction. When many signals are summed and are randomly distributed in both frequency and phase, the ratio of peak-to-average voltage rarely exceeds 14 dB (though with occasional higher peaks). In a CATV system, however, the downstream spectrum is not random. Peak voltage conditions may occur, for example, when a large number of carriers are harmonics of a common root frequency and the carrier phases are aligned. In that case, the time domain waveform can resemble a string of impulses spaced by a time interval equal to the period of the common root frequency. As a result of this occasionally occurring peak voltage (and thus peak drive current) condition, the laser may be driven into hard limiting, causing clipping, when a sufficient number of carriers are in phase alignment. This is particularly true in the case of directly modulated laser diodes, as described above, where a sharp knee occurs in the transfer function below which the light output reaches a zero-power level.
In other words, there will be clipping when the instantaneous sum of various signals causes the drive current to swing too far in the “downward” direction and below the threshold current that turns on the laser. When such clipping occurs, intermodulation products (i.e., clipping-induced distortion) and noise may be generated, which may result in bit errors in the optical output of the laser. Systems including a digital signal (e.g., a QAM signal) added to the analog channels are even more sensitive to noise arising from clipping.
Each channel in a multichannel optical communications system may be driven or modulated up to a certain maximum optical modulation index (OMI). In general, a higher OMI per channel increases the channel-to-noise ratio (CNR). Driving the individual channels too much, however, may cause clipping when the channels line up, as described above. In some systems, each channel cannot be driven more than about 2 to 3% OMI because, if the channels were to line up, the total modulation of the laser would substantially exceed 100%. If clipping could be reduced or eliminated, increasing the OMI per channel would be possible to improve CNR.
Some systems for reducing or preventing clipping have used peak detection circuits to identify amplitude spikes that may result in clipping. When such peaks are detected, the RF drive signal is delayed and the laser bias is temporarily increased until the clipping event has passed. In such systems, however, the long RF delay between the peak detector circuit and the laser transmitter may be difficult to implement.
Other systems for reducing or preventing clipping have used peak detectors to detect peaks, to predict the subsequent peaks and to increase the bias current to reduce or prevent clipping in the subsequent peaks. One problem with this approach includes the difficulty of precisely setting the adjustment window so that it coincides exactly with the subsequent clipping events.
Other systems for reducing or preventing clipping have upconverted the RF signal, detected the envelope of the upconverted signal and used the detected envelope to adjust the laser bias. In such systems, however, an upconverter is required and the envelope detector is specifically designed for detecting an envelope of the upconverted RF signal.